While I see the gist of what you are saying, and the rumour / credibility thing is realistically always an issue in such cases AFAIK, I do not think it is true that "Free speech only exists with attribution [...] you advance nothing in an environment of anonymity."
If a new space for anonymous speech comes up, things will be chaotic at first, trust chains and circles - 'web(s) of trust' (not sure of terminology heh) do emerge, and I've seen it happen [citation neeeded]. Consider the Bitcoin over-the-counter marketplace (#bitcoin-otc on Freenode), where a web of trust (based on (potentially) anonymous PGP identities/keypairs) does function quite well (not without failures). Actually, if the identities are tied to something like a PGP keypair, it obviously works across (e.g.) forum boards (cough (Tor, etc.) underground forum scene cough). You could actually use PGP signature chaining, etc. (it does work very nicely!)
Of course, in my (vague) illustration, it'd be much more chaotic and nasty. The thing is though that in the end, people do sense a need to have a medium to coordinate efforts, exchange info, etc. (Consider also e.g. the idea that I can disclose my real identity to a select party (pre-arranged IRL, e.g.), but not necessarily to the whole forum. However, if that party is trusted by other nodes, then those nodes can trust me without knowing who I am. Lots of human factors and points of failure here, though. But it is not always futile!) At the very least, one could coordinate an IRL meeting (you would of course say, what if the organizers are covert government agents, etc.) In the end, a system connected to IRL matters and lives will have IRL-bound points of failure. That does not mean that it could (or does) not work, or that it would be as fallible as a non-(quasi-)anonymous solution.
But I agree that it's usually a lot of effort; not necessarily futile though, and that's my only point really.
If a new space for anonymous speech comes up, things will be chaotic at first, trust chains and circles - 'web(s) of trust' (not sure of terminology heh) do emerge, and I've seen it happen [citation neeeded]. Consider the Bitcoin over-the-counter marketplace (#bitcoin-otc on Freenode), where a web of trust (based on (potentially) anonymous PGP identities/keypairs) does function quite well (not without failures). Actually, if the identities are tied to something like a PGP keypair, it obviously works across (e.g.) forum boards (cough (Tor, etc.) underground forum scene cough). You could actually use PGP signature chaining, etc. (it does work very nicely!)
Of course, in my (vague) illustration, it'd be much more chaotic and nasty. The thing is though that in the end, people do sense a need to have a medium to coordinate efforts, exchange info, etc. (Consider also e.g. the idea that I can disclose my real identity to a select party (pre-arranged IRL, e.g.), but not necessarily to the whole forum. However, if that party is trusted by other nodes, then those nodes can trust me without knowing who I am. Lots of human factors and points of failure here, though. But it is not always futile!) At the very least, one could coordinate an IRL meeting (you would of course say, what if the organizers are covert government agents, etc.) In the end, a system connected to IRL matters and lives will have IRL-bound points of failure. That does not mean that it could (or does) not work, or that it would be as fallible as a non-(quasi-)anonymous solution.
But I agree that it's usually a lot of effort; not necessarily futile though, and that's my only point really.